In some solar thermal power plants, numerous heliostats may be employed to reflect light onto a receiver. The mirrors of each of the heliostats must be continually repositioned in order to account for the movement of the sun. Tracking errors must be exceedingly small in large power plants to achieve high concentration at the receiver aperture. When the heliostats are installed, however, the precision with which the location and orientation of the heliostats is known is generally insufficient to accurately reflect light to the receiver. A calibration phase is then necessary to generate better estimates of the position and orientation variables.
For calibration, power plants generally use one of several systems: a white screen located near the receiver aperture with one or more cameras pointed at the receiver, or multiple cameras located near the aperture looking onto the heliostats. In the first system, each heliostat redirects light to one or more white screens located in the vicinity of the receiver. An external camera is used to locate the reflected light of said heliostat and determine the alignment error based upon the difference between expected and measured positions. Only a small number of heliostats can be calibrated at a time since their reflected images are projected onto the same white screen. Disadvantages of this system include (a) too long a calibration time since a large number of mirrors must be calibrated serially; (b) incomplete characterization of the heliostat pose since the error signal is essentially on the receiver plane, (c) dependence on centralized control and connectivity with the latter, (d) open-loop operation after calibration which is not robust with respect to shifts in heliostats' base coordinates and orientation (e.g., due to land shifts, earthquakes, etc).
In a second prior art system, multiple cameras mounted near the receiver control the heliostat in closed-loop. In one implementation, four cameras are positioned at the receiver, one to the right, above, left, and below the receiver aperture. Each camera is pointed at the heliostat field; since the optics of each camera is close enough to a pinhole, distinct heliostats pointed at a given camera will be imaged as a distinct bright spot in the camera plane. Thus several heliostats can be imaged in parallel, overcoming the serial limitation of the white-screen based prior art. Heliostats can thus be aimed in closed-loop, namely, the cameras can guide a given heliostat to aim at the receiver aperture exactly between them. This system, however, poses several practical difficulties. First, the surrounding cameras must lie exceedingly close to the receiver aperture, thus increasing the chance of damage to the cameras if exposed to concentrated flux. Second, each camera must be able to image the entire field (requiring very large field of view) and resolve all heliostats (requiring very fine resolution), which is especially difficult for fields with a very large number of small heliostats.
There is therefore a need for a cost-effective, practical, sensor-robust, and decentralized heliostat tracking system that permits each of a large array of heliostats to accurately reflect sunlight to the receiver, especially where the heliostat mirrors are densely populated and at large distances from the receiver.